HealthRAG is a Retrieval-Augmented Generation (RAG) chatbot designed for medical document question-answering, built with Python, LangChain, FAISS, and Hugging Face API.

HealthRAG is a medical Q&A chatbot that responds to user queries by retrieving relevant information from PDF documents. It is intended for:

• 🩺 Patients seeking general medical insights (may contain technical language)
• 👨‍⚕️ Medical students for revision, summarization, and context-based querying

The quality of answers depends heavily on the PDFs provided as context.

1. Users upload one or more PDF files containing medical content
2. The content is split into chunks and converted into embeddings
3. FAISS is used to store the embeddings and retrieve the most relevant context
4. A custom prompt sends the context + user question to a hosted LLM via Hugging Face Router
5. The chatbot responds concisely using only the retrieved context

git clone https://github.com/Y1D1R/HealthRAG.git cd HealthRAG

Create a .env file in the root with:

HF_TOKEN=your_huggingface_access_token HF_API_TOKEN=your_huggingface_api_token

👉 You can get this from https://huggingface.co/settings/tokens

pip install -e .

Place your medical .pdf files in the data/ directory.

python app/components/data_loader.py

python app/application.py

Visit the app at: http://127.0.0.1:5000

Place your PDF documents in the local data/ folder before building the image.

Ensure your .env file with API keys is available in the root before the build.

docker build -t healthrag .

✅ During the build, the vector store will be automatically generated from the contents of data/.

docker run -p 5000:5000 --name healthrag-container healthrag

Then visit the app at: http://localhost:5000

HealthRAG/ ├── app/ │ ├── components/ # pdf_loader, embeddings, vector store │ ├── config/ # API keys, chunk config │ ├── common/ # Logger, custom exception │ └── application.py # Flask web interface ├── templates/ # Frontend HTML ├── static/images/ # Logo ├── data/ # Your medical PDFs ├── Dockerfile ├── setup.py ├── Jenkinsfile └── README.md 

• Python 3.11
• Flask
• LangChain
• Hugging Face Hub & Router
• FAISS (vector store)
• Sentence Transformers (embeddings)
• OpenAI-compatible openai Python client

This chatbot is a proof of concept and not intended for clinical use. Its accuracy depends solely on the content of the provided PDFs.